PSA screening and mortality from prostate cancer.

B4

Introduction: Prostate cancer is the 2^nd-leading cause of cancer death among men and is particularly common among elderly men. Screening for prostate cancer with the prostate specific antigen (PSA) test continues to be a contentious issue. Findings support the PSA test's ability to detect early-stage disease, but no prospective study has investigated its benefit in the context of screening. Three observational studies have been published. One by Kopec et al. claimed a significant benefit that was only apparent in multivariable analysis (OR of 0.65). The study by Concato et al. with subjects from the Veteran's Administration showed no benefit (OR of 1.08), and a third study by Weinmann et al. demonstrated protection from the digital rectal exam, but did not isolate any PSA effect. We now report the results of a new observational study of PSA screening. Methods: This was a case-control study using fatal cases ascertained from New Jersey death certificates. Population-based controls were individually matched to cases on race, age, and time for possible exposure to a PSA test as indexed by the date of suspected diagnosis in the case. Possible confounders (comorbidity and education level) were adjusted in conditional logistic regression models. Results: The frequency of screening in controls was very similar to that reported elsewhere for New Jersey males. Controls were moderately more educated than our cases but were similar for age, race, and number of comorbidities. The odds of dying having been screened with a PSA vs. not being screened was 1.07, 95% CI of 0.74 to 1.56 in our unadjusted analysis (conditional logistic regression accounting for the matching variables only). After adding educational level and comorbidity number, the adjusted odds ratio (AOR) was 1.03, 95% CI 0.70 to 1.53). We repeated the analysis using various restrictions for what we considered to be a "screening" PSA, and the results were quite similar. Conclusions: Our study using a population-based source of cases and controls does not show PSA screening to offer any protection from prostate cancer mortality. Possible limitations are the time frame for exposure to a PSA test and the possibility of missed PSA screens from the medical records. Strengths include the diversity of the population examined and the inclusion of sensitivity analyses of different criteria for identifying PSA "screens". This study adds to the knowledge base of the other three observational studies done for this screening modality. There are now two null studies for PSA testing, and one showing protection. Although there may be a benefit of PSA testing in certain subgroups, these results should remind clinicians to be circumspect and thoughtful when ordering a PSA test in the context of screening. We, as others, await the results of the two randomized trials.